In this talk, I conceptualize and evaluate how using large language models (LLMs) transforms our writing practices as well as our cognition and critical thinking skills. Drawing on empirical research, I outline how people use LLMs for writing tasks and how this transforms their writing practices, cognition, and critical thinking skills. This shows that LLMs can (1) hinder the development of writing skills in students and (2) when certain writing skills have been learned in the past, those might be reduced due to lack of practice. Next, I reflect on the desirability of these cognitive transformations, arguing that the cognitive benefits of task efficiency might not outweigh the drawbacks of cognitive deskilling. Finally, I outline why we should cultivate writing practices involving pen and paper or a word processor. It does so by conceptualizing how writing and text create a material cognitive workspace with informational and functional properties that complement those of the unaided embodied brain, affording operations and manipulations that are very difficult, if not impossible, to perform internally. Writing is cognitively advantageous because it enlarges the space of possible thoughts, insights, and knowledge, also generating positive epistemic feelings of reward and accomplishment.
Author: Richard Heersmink